Re: [AMBER] Problem with

From: Jason Swails <>
Date: Tue, 8 Jul 2014 22:50:50 -0700

On Tue, Jul 8, 2014 at 10:32 PM, Rajeswari A. <>

> Hi Jason,
> Thank you for your reply. As you mention, sander and MMPBSA_py_energy agree
> well for the same input with same trajectory. When I did a run with
> decomposition, mmpbsa used sander by default and gave me EPB of -41. Same
> input without decomposition ran in MMPBSA_py_energy by default and gave me
> EPB = -2109. I misunderstood that the difference might be due to the PBSA
> solver. As per your suggestion i ran another set without decomposition and
> used sander for solving PBSA. That agree well with the mmpbsa_py_energy.
> Now i did not know why the run with additional decomposition analysis for
> the same trajectory showed high degree of variation. For your reference I
> am providing here the output files of two runs. Please suggest me what is
> going wrong..

I have no idea how pairwise energy decomposition is done using PB. What I
_can_ tell you, however, is that it is impossible to do "right", since
energies are not pairwise decomposable. This is true for the very simple
reason that the dielectric boundary is defined by every atom in the system,
which has a direct impact on the interaction between residues. I'm unsure
what approximation is made in this case, but it would appear to destroy the
estimate of the total solvation free energy.

The GB formalism is more amenable to pairwise decomposition, since the
energy involves summing over pairs of particles. However, even with GB the
energies are not truly pairwise decomposable since the sum-over-pairs
includes the effective Born radii for each of the atoms in question which
are computed as an integral over all other atoms in the system. Therefore,
the energy between two atoms in GB still depends on other atoms in the
system. At least with GB, though, the final solvation free energy is the
same whether decomposition is used or not.

I guess what I'm trying to say is that the moral of this story is pairwise
energy decomposition is at best a qualitatively useful tool that can point
out some hot-spot residues worth targeting for mutations or drug design. I
would certainly not rely on it for any kind of quantitative analysis.


P.S., In my experience, though, decomposition usually just points out the
obvious. "This arginine residue close to the binding site for a
carboxylate ligand has a far greater effect than other, neutral residues
surrounding it!" <--You didn't need to predict that :).

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Tue Jul 08 2014 - 23:00:04 PDT
Custom Search